Design of a Graphical User Interface for rapid identification and parameterisation of bony landmarks of scoliotic spines from CT scans

Abstract

Introduction
Spinal deformity disfigures the torso and reduces lung capacity. Consequently, it has a major detrimental effect on both the physical and psychological wellbeing of its sufferers. The main goal of this project is to improve the outcome of surgical deformity correction by providing patient-specific Finite Element (FE) models derived from pre-operative CT scans to predict the effects of surgical procedures.

Methods
The FE models developed in this project are parametric, meaning they are not derived by directly meshing CT data, but by detecting bony landmarks from CT data to which a predefined mesh is fitted. Landmarks are identified using a custom-developed Graphical User Interface (GUI) which contains algorithms for automatic detection of specific groups of landmarks such as ribs and endplates. Landmark properties are fed directly into a preprocessor that generates the FE mesh.

Results
Identifying all required landmarks of a complete spine takes between one and two hours. Intraobserver variations in coordinates and angles are small (standard deviations usually below 2mm and 3deg respectively).

Discussion & Conclusion
Many of the commercially available algorithms can derive FE meshes directly from CT data in seconds, but high mesh densities and tetrahedral elements make these models unsuitable for whole-spine FE simulations. The semi-automated preprocessor presented allows better control over the FE mesh and material property assignment, and should eventually lead to a process where the surgeon is provided with a reliable FE prediction of the biomechanical outcome of the proposed surgery within a day or two of taking the CT scans.